Symmetric matrices and positive deniteness

Symmetric matrices are good their eigenvalues are real and each has a com
plete set of orthonormal eigenvectors. Positive denite matrices are even bet
ter.

Symmetric matrices
A symmetric matrix is one for which A = A T . If a matrix has some special
property (e.g. its a Markov matrix), its eigenvalues and eigenvectors are likely
to have special properties as well. For a symmetric matrix with real number
entries, the eigenvalues are real numbers and its possible to choose a complete
set of eigenvectors that are perpendicular (or even orthonormal).
If A has n independent eigenvectors we can write A = SS1 . If A is sym
metric we can write A = QQ1 = QQ T , where Q is an orthogonal matrix.
Mathematicians call this the spectral theorem and think of the eigenvalues as the
spectrum of the matrix. In mechanics its called the principal axis theorem.
In addition, any matrix of the form QQ T will be symmetric.
Real eigenvalues
Why are the eigenvalues of a symmetric matrix real? Suppose A is symmetric
and Ax = x. Then we can conjugate to get Ax = x. If the entries of A
are real, this becomes Ax = x. (This proves that complex eigenvalues of real
valued matrices come in conjugate pairs.)
Now transpose to get x T A T = x T . Because A is symmetric we now have
T
x A = x T . Multiplying both sides of this equation on the right by x gives:
x T Ax = x T x.
On the other hand, we can multiply Ax = x on the left by x T to get:
x T Ax = x T x.
Comparing the two equations we see that x T x = x T x and, unless x T x is zero,
we can conclude = is real.
How do we know x T x = 0?

x1

x2
x T x = x1 x2 x n . =
|
x1 |
2 +
|
x2 |
2 +
+
|
xn |
2 .
..
xn
If x = 0 then x T x = 0.

With complex vectors, as with complex numbers, multiplying by the conju

gate is often helpful.
Symmetric matrices with real entries have A = A T , real eigenvalues, and
perpendicular eigenvectors. If A has complex entries, then it will have real
T

eigenvalues and perpendicular eigenvectors if and only if A = A . (The proof

of this follows the same pattern.)
Projection onto eigenvectors
If A = A T , we can write:
A

= QQ T

q1

q2

qn

1
2
..

.
n

q1T
q2T
..
.

qnT

= 1 q1 q1T + 2 q2 q2T + + n qn qnT

The matrix qk qkT is the projection matrix onto qk , so every symmetric matrix is
a combination of perpendicular projection matrices.
Information about eigenvalues
If we know that eigenvalues are real, we can ask whether they are positive or
negative. (Remember that the signs of the eigenvalues are important in solving
systems of differential equations.)
For very large matrices A, its impractical to compute eigenvalues by solv
ing | A I | = 0. However, its not hard to compute the pivots, and the signs
of the pivots of a symmetric matrix are the same as the signs of the eigenvalues:
number of positive pivots = number of positive eigenvalues.
Because the eigenvalues of A + bI are just b more than the eigenvalues of
A, we can use this fact to nd which eigenvalues of a symmetric matrix are
greater or less than any real number b. This tells us a lot about the eigenvalues
of A even if we cant compute them directly.

Positive denite matrices

A positive denite matrix is a symmetric matrix A for which all eigenvalues are
positive. A good way to tell if a matrix is positive denite is to check that all
its pivots are positive.


Let A =

5
2

2
3

. The pivots of this matrix are 5 and (det A)/5 = 11/5.

The matrix is symmetric and its pivots (and therefore eigenvalues) are positive,
so A is a positive denite matrix. Its eigenvalues are the solutions to:

| A I | = 2 8 + 11 = 0,

i.e. 4 5.
The determinant
of a positive denite matrix is always positive but the de

terminant of

1
0

0
3

is also positive, and that matrix isnt positive de

nite. If all of the subdeterminants of A are positive (determinants of the k by

k matrices in the upper left corner of A, where 1 k n), then A is positive
denite.
The subject of positive denite matrices brings together what weve learned
about pivots, determinants and eigenvalues of square matrices. Soon well
have a chance to bring together what weve learned in this course and apply it
to non-square matrices.

MIT OpenCourseWare
http://ocw.mit.edu

18.06SC Linear Algebra

Fall 2011

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.